Abstract
Abstract
DNA methyltransferase inhibitor drugs (MTIs) such as decitabine can overcome gene silencing due to aberrant hypermethylation of gene promoters. Presumably, this effect is responsible for the therapeutic activity of MTIs as clinically demonstrated in myelodysplasias (MDS) and leukemias. Other tumors such as diffuse large B-cell lymphomas (DLBCLs) can also present with aberrant promoter hypermethylation. However, it is currently difficult to prospectively identify patients likely to respond to MTIs, since specific methylation markers or signatures have not yet been identified. We predicted that decitabine would have anti-lymphoma activity in a subset of DLBCLs, and that these cases would exhibit specific methylation signatures predictive of response to these drugs. To determine whether this is the case we first exposed a panel of 7 DLBCL cell lines (Ly1, Ly7, Ly10, SU-DHL6, Farage, Pfeiffer and Toledo) to increasing concentrations of decitabine (0.5, 1, 2.5, 5, 10, 50 and 100 μM) administered after synchronization by 12 hr serum starvation. Viability was assessed after 48 hr of culture by MTS-based assay and Trypan blue exclusion. The IC25 and IC50 were calculated for all cell lines by constructing dose-response curves. The IC25 was used to discriminate sensitive (6.3 ± 1.2 μM) vs. resistant (49.4 ± 5 μM, p < 0.01) cell lines. Interestingly, there was no correlation between MTI sensitivity and DLBCL subtype as defined by recent gene expression profiling classification efforts (i.e. GCB vs. ABC, or BCR vs. OxPhos). To identify the methylation signatures of these DLBCL cells we used a method that we developed for genome-wide DNA methylation quantification called HELP (HpaII tiny fragment Enrichment by LM-PCR). HELP is based on comparative Msp1 and HpaII digestion of genomic DNA, followed by size specific amplification and co-hybridization to custom high-density oligonucleotide arrays designed to provide uniform data collection over 25,000 promoters. HELP compares favorably to other high throughput methods in that it is highly reproducible (R > 0.98) and has an extremely robust signal-to-noise ratio. DNA was collected from the DLBCL cells for HELP prior to drug treatment. Most significantly we found that unsupervised (i.e. unbiased) clustering of DNA methylation profiles could readily segregate decitabine resistant vs. sensitive DLBCL cell lines. Correspondence analysis clearly identified a methylation signature consisting of 133 differentially methylated genes that distinguishes between decitabine sensitive and resistant cells. Most of these appeared to be functionally relevant including such genes as Caspase-9, RARB, JUNB, and ELK1. Biological assays to determine the contribution of these genes to the phenotype are underway. Taken together, our data suggest that MTIs might be effective in a cohort of DLBCL cases that exhibit the specific methylation signature that we have identified. Prospective evaluation of the predictive value of this signature may allow optimal selection of patients for clinical trials with these agents.